#AI #LLM #Qwen #Claude <details class="toc-container"> <summary><strong>Table of Contents</strong></summary> <ul> <li>Introduction</li> <li>Model Comparison Tests</li> <ul> <li>Weather Card Design Test</li> <li>Animated Weather Card Implementation</li> </ul> <li>Technical Notes</li> </ul> </details> ## Introduction On February 25th, 2025, Anthropic released its new model `claude-3-7-sonnet`, while Alibaba unveiled an enhanced reasoning model based on their `Qwen2.5-Max` architecture (initially released in January 2025). The `Qwen2.5-Max` release was overshadowed by `deepseek-r1` at the time, but Alibaba has announced plans to open-source both their base model and the new reasoning variant. This document presents comparative testing results between these models. ## Model Comparison Tests ### Weather Card Design Test 1. The idea of the first test case is from [Yoshiki Miura](https://x.com/miiura) and I do some modifications. My test prompt is: "Make a more modern design for the weather card in Shanghai". #### Claude-3-7-Sonnet Results - **Without Thinking Mode:** ![[claude-3-7-nothinking.png]] - **With Thinking Mode:** ![[claude-3-7-thinking.png]] *Note: Results were similar but more concise with thinking enabled* #### Qwen2.5-Max Results - **Without Thinking Mode:** ![[qwen2-5-max.png]] - **With Thinking Mode:** ![[qwen2-5-max-thinking.png]] ### Animated Weather Card Implementation 2. The idea of my second case is from X user [低空飞行](https://x.com/localhost_4173). The test prompt is more complicated: ```PlainText Creating a single HTML file containing CSS and JavaScript to generate an animated weather card. The card should visually represent the following weather conditions with distinct animations: Wind:(moving clouds, swaying trees or wind lines) Rain:(e.g. falling raindrops, puddings forming); Sun(e.g., shining rays, bright background) Snow:(e.g, falling snowflakes, snow accumulating) Show all the weather card side by side. The card should have a dark background. Provide all the necessary HTML, CSS, and JavaScript. ``` #### Claude-3-7-Sonnet Results - **Without Thinking Mode:** ![[claude-3-7-nothining-animateweather.png]] - **With Thinking Mode:** ![[claude-3-7-thining-animateweather.png]] #### Qwen2.5-Max Results - **Without Thinking Mode:** ![[qwen2-5-max-nothinking-animated-weather.png]] - **With Thinking Mode:** ![[qwen2-5-max-thinking-animateweather.png]] ## Technical Notes - **Claude-3-7-Sonnet API Features:** - Configurable thinking budget (up to 128k tokens) - Allows optimization of speed/cost vs. answer quality - Thinking tokens can be precisely controlled - **Claude Code Release:** - New coding-focused tool from Anthropic - Key capabilities: - Code search and analysis - File editing - Test writing and execution - GitHub integration - Command line tool support - Similar functionality to the open-source `aider` tool